Semiconductor devices such as memories typically have conductive links adhered to a transparent insulator layer such as silicon oxide, which is supported by the main silicon substrate. During laser processing of such semiconductor devices, while the beam is incident on the link or circuit element, some of the energy also reaches the substrate and other structures. Depending upon the power of the beam, length of time of application of the beam, and other operating parameters, the silicon substrate and/or adjacent can be overheated and damaged.
Several prior art references teach the importance of wavelength selection as a critical parameter for substrate damage control. U.S. Pat. Nos. 4,399,345, 5,265,114, 5,473,624, 5,569,398 disclose the benefits of wavelength selection in the range beyond 1.2 um to avoid damaging silicon substrates.
The disclosure of the above-noted '759 patent further elaborates on the wavelength characteristics of silicon. The absorption in silicon rapidly drops off after about one micron with an absorption edge of about 1.12 microns at room temperature. At wavelengths greater than 1.12 microns, the silicon starts to transmit more and more easily and, thus, it is possible to obtain better part yields upon removing material from the silicon. In the range around 1 micron the absorption coefficient decrease by a factor of four orders of magnitude going from 0.9 microns to 1.2 microns. In going from the standard laser wavelength of 1.047 microns to 1.2 microns the curve shows a drop of two orders of magnitude. This shows a drastic change in absorption for a very slight change in wavelength. Thus, operating the laser at a wavelength beyond the absorption edge of the substrate circumvents damage to the substrate, which is especially important if there is a slight misalignment of the laser beam with respect to the link or where the focused spot extends beyond the link structure. Furthermore, if the substrate temperature rises during processing the absorption curve shifts will shift further into the infrared which can lead to thermal runaway conditions and catastrophic damage.
The problem of liquid crystal repair is similar to the problem of metal link ablation. The wavelength selection principle for maximizing absorption contrast was advantageously applied in the green wavelength region in a manner analogous to the above disclosures for the same purpose—namely removal of metal without substrate damage. The system manufactured by Florod is described in the publication “Xenon Laser Repairs Liquid Crystal Displays”, LASERS AND OPTRONICS, pages 39-41, April 1988.
Just as wavelength selection has proven to be advantageous, it has been recognized that other parameters can be adjusted to improve the laser processing window. For example, it was noted in “Computer Simulation of Target Link Explosion in Laser Programmable Redundancy for Silicon Memory” by L. M. Scarfone and J. D. Chlipala, 1986, p. 371, “It is desirable that laser wavelengths and various material thicknesses be selected to enhance the absorption for the link removal process and reduce it elsewhere to prevent damage to the remainder of the structure.” The usefulness, in general, of thicker insulative layers underneath links or circuit elements and the usefulness of limiting the duration of heating pulses has also been recognized, as in the paper co-authored by the applicant, “Laser Adjustment of Linear Monolithic Circuits,” Litwin and Smart, 100/L.I.A., Vol. 38, ICAELO (1983).
The '759 patent teaches the tradeoffs that exist with selection of the longer wavelengths—specifically compromises with respect to spot size, depth of focus, and pulse width, available from Nd:YAG lasers. These parameters are of critical importance for laser processing at increasingly fine dimensions, and where the chances of collateral damage to surrounding structures exist.
In fact, any improvement which widens the processing window is advantageous as the industry continues to push toward higher density microstructures and the associated geometries which are a fraction of one micron in depth or lateral dimension. The tolerances of energy control and target absorption become large compared to the energy required to process the microstructure at this scale. It should be noted from the above discussion that laser processing parameters are not necessarily independent in micromachining applications where a small laser spot, about 1 μm, is required. For instance, the spot size and pulse width are generally minimized with short wavelengths, say less than 1.2 μm, but the absorption contrast is not maximized. Makers of semiconductor devices typically continue production of earlier developed products while developing and entering production of more advanced versions that typically employ different structures and processes. Many current memory products employ polysilicide or polysilicon links while smaller link structures of metal are used for more advanced products such as the 256-megabit memories. Links of 1 micron width, and ⅓ micron depth, lying upon a thin silicon oxide layer of 0.3 to 0.5 microns are being used in such large memories. Production facilities traditionally have utilized Q-switched diode pumped YAG lasers at and related equipment capable of operating at the conventional wavelengths of 1.047 μm-1.32 μm and related equipment capable of operating in the wavelength region recognized for its lower absorption by silicon. However, these users also recognize the benefits of equipment improvements which results in clean severing of link structures without the risk of later chip failures due to conductive residue or contamination near the ablation site.
Other degrees of freedom include laser pulse energy density (delivered to the target) and pulse duration. It has been taught in the prior art that pulse width should be limited to avoid damage in micromachining applications. For example, in U.S. Pat. No. 5,059,764 a laser processing workstation is disclosed wherein a q-switched laser system is utilized to produce, among other things, relatively short pulses on the order of 10-50 ns. It was disclosed that for material processing applications (like semiconductor memory repair via link blowing and precision engraving), the output pulse width should be relatively short—and that a pulse width less than 50 ns is required in many applications, for example 30 ns. The proper choice of pulse width allows for ablation (evaporation without melting).
High speed pulsed laser designs may utilize Q-switched, gain switched, or mode-locked operation. The pulse duration and shape of standard Q-switched and other pulsed lasers can be approximated at a fundamental level by integrating the coupled rate equations describing the population inversion and the photon number density relative to the lasing threshold at the start of the pulse. For the Q-switched case, on a normalized scale, a higher number of atoms in the inverted population relative to the threshold the faster the pulse rise time, the narrower the width, and the higher the peak energy. As the ratio decreases the pulse shape becomes broader with lower energy concentration.
Often Q-switched laser pulses resemble a Gaussian temporal distribution, or a mixture of a Gaussian with an exponential decaying tail. As disclosed in the '759 patent, the shorter wavelength diode pumped systems are capable of producing relatively short pulses, about 10 ns, when measured at the half power points (i.e., standard definition of pulse duration) and are operated in a favorable wavelength region. Despite successful operation, applicant has found several limitations associated with the temporal pulse shape characteristic of standard diode pumped Q-switch laser systems, including the practical rise time limitations, the power distribution between the half maximum points, and the pulse decay characteristic which, when improved using the method and system of the present invention, provided noticeably better results in a metal link blowing application.
Throughout the remainder of this specification, “pulse shaping” refers to the generation of a laser pulse which is to be detected with a detector of electromagnetic radiation where “shape” refers to the power on the detector as a function of time. Furthermore, “pulse width” or “pulse duration” refers to the full width at half maximum (FWHM) unless otherwise stated. Also, Q-switched pulses collectively refers to temporal distribution of pulses obtained, for example, in standard Q-switched systems which may resemble a mixture of a substantially Gaussian central lobe with a relatively slow decaying exponential tail. These wave shapes are formally referred to as a “Q-switched pulse envelope” in laser literature. FIG. 1c shows such pulses.
In U.S. Pat. No. 5,208,437 (i.e., the '437 patent), a pulse width specification of less than 1 ns was specified for a memory repair application. Earlier work by the co-inventors of the '437 patent disclosed in “Laser Cutting of Aluminum Thin Film With No Damage to Under Layers”, ANNALS OF THE CIRP, Vol 28/1, 1979, included experimental results with relatively short laser pulses having a “Gaussian” shape as defined above. The results indicated a “desired portion of the interconnection pattern” which is made of aluminum or the like, “can be cut without the layer disposed below the interconnection pattern being damaged”. Specifications for the pulse width of substantially 1 ns or less with energy density of substantially 106 W/cm2 were disclosed for the apparatus. However, there was no disclosure regarding a method of temporal pulse shaping, although spatially the beam was shaped to correspond to the interconnection pattern. Furthermore, applicant's analysis on high density memory devices having multiple layers with specified pulsewidths in the ultrafast range, which is approached with the 100-300 ps used in the '437 patent, have not been satisfactory. Overcoming this limitation would presently require the ultrafast laser system to produce multiple pulses for processing each target site which would slow the laser processing rate to an unacceptable level.
Continuing to the ultrafast scale, experimental results have been disclosed for micromachining operations. The ultrafast pulses have durations on the order of fs (10-15 sec) to ps (10-12) and, at the decreased scale, exploit material properties at the atomic and molecular which are fundamentally different than found in the range of several hundred ps to ns.
In U.S. Pat. No. 5,656,186 and the publication “Ultrashort Laser Pulses tackle precision Machining”, LASER FOCUS WORLD, August 1997, pages 101-118, machining operations at several wavelengths were analyzed, and machined feature sizes significantly smaller than the diffraction limited spot size of the focused beam were demonstrated.
Laser systems for ultrafast pulse generation vary in complexity and are exemplary embodiments are described in U.S. Pat. Nos. 5,920,668 and 5,400,350, and in Ultrafast Lasers Escape The Lab”, PHOTONICS SPECTRA, July 1998, pp. 157-161. The embodiments generally include methods for pulse stretching of mode locked ultrafast pulses prior to amplification to avoid amplifier saturation followed by compression to extremely narrow widths. This technology holds promise for certain class of micromachining and possibly finer scale “nanomachining” operations, the latter benefit afforded by machining below diffraction limit. However, Applicant has discovered practical limitations at the present time with the available power in each pulse for applications like metal link blowing and similar micromachining applications leading to the unacceptable requirement for multiple pulses.
Applicant wishes to elaborate on the rationale for the use of a short pulse, fast rise time pulse is indicated in the following paragraphs as the reasons are manifold and a number of theoretical and empirical papers and books have been written on the subject. Ablation of metal links is taken as an example, although the principles extend to many laser processing applications where a target material is surrounded by material having substantially different optical and thermal properties. The following references 1-3 are examples:                1. John F. Ready, Effects of High Power Laser Radiation, ACADEMIC PRESS, New York 1971, pages 115-116.        2. Sidney S. Charschan, Guide for Material Processing By Lasers, Laser Institute of America, The Paul M. Harrod Company, Baltimore Md., 1977, pages 5-13.        3. Joseph Bernstein, J. H. Lee, Gang Yang, Tariq A. Dahmas, Analysis of Laser Metal-Cut Energy Process Window (to be published).Metal Reflectivity        
Metal reflectivity decreases with increased power density of a laser pulse (ref. 1). The reflectivity of a metal is directly proportional to the free electron conductivity in a material. At high electric field densities as delivered by a high intensity laser, the collision time between electrons and the lattice is reduced. This shortening of the collision time reduces the conductivity and hence the reflectivity. For example, the reflectivity of aluminum decreases from 92% to less than 25% as the laser power densities increases to the range of 109 watts/cm2. Hence, to circumvent the loss of laser energy to reflection it is advantageous to achieve high power density at the work piece in as short a time as possible.
Thermal Diffusivity
The distance D that heat travels during a laser pulse is proportional to the laser pulse width as follows:D=√{square root over (kt)}where:    K is the thermal diffusivity of the material; and    t is the length of the laser pulse.
Hence, it can be seen that a short laser pulse prevents heat dissipating to the substrate below the melting link and also heat conducting laterally to the material contiguous to the link. However the pulse must be long enough to heat the link material all the way through.
Thermal Stress and Link Removal
Through the absorption of the laser energy the target metal link heats up and tries to expand. However, the oxide surrounding the link contains the expanding material. Hence, stress is built up within the oxide. At some point the pressure of the expanding metal exceeds the yield point of the oxide and the oxide cracks and the metal link explodes into a fine particle vapor. The principal crack points of metal link occurs at the maximum stress points, which are at the edges of the link both top and bottom as shown in FIG. 1b. 
If the oxide over the link is somewhat thin then the cracking of the oxide will occur at the top of the link only and the oxide and link will be removed cleanly as shown in FIG. 1a. However, if the oxide is somewhat thick, cracking can occur at the bottom of the link as well as the top and the crack will propagate down to the substrate as shown in FIG. 1b. This is a highly undesirable circumstance.
Q-switched laser systems can be modified to provide short pulses of various shapes. Typical prior art lasers that produce high peak power, short pulse lasers are standard Q-switched lasers. These lasers produce a temporal pulse having a moderate pulse rise time. It is possible to change this temporal shape by using a Pockels Cell pulse slicer that switch out sections of the laser beam. In U.S. Pat. No. 4,483,005 (i.e., the '005 patent), invented by the Applicant of the present invention and having the same assignee, various methods for affecting (i.e., reducing) laser beam pulse width are disclosed. As taught in the '005 patent, which is hereby incorporated by reference, the laser pulse can be shaped somewhat to produce a “non-Gaussian” shaped beam by truncating energy outside the central lobe. It should be noted that if a relatively broad Q-switched waveform is to be transformed to a narrow, uniform shape, only a small fraction of the pulse energy will be used. For example, truncation of a Gaussian pulse to provide a sharp rise time and a narrow pulse with flatness to within 10% reduces the pulse energy by about 65%.
Similarly, in U.S. Pat. No. 4,114,018 (the '018 patent), temporal pulse shaping to produce square pulses is disclosed. FIG. 7 shows the time interval for relatively flat laser power output. In the '018 patented method, it is necessary to remove a temporal segment of the beam intensity in order to generate the desired pulses.
A desirable improvement over the prior art would provide an efficient method for generating short pulses with high energy enclosure within the pulse duration with rapidly decaying tails. In order to accomplish this, laser technology which produces pulse shapes different than those of the Q-switched pulse envelope is preferred. Such pulses have fast rise time, uniform energy in the central lobe, and fast decay.
The fast rise-time, high power density pulse as produced by a laser other than a standard Q-switched Nd:YAG will best accomplish this task.
These benefits are implemented in a preferred manner in a system which uses laser technology departing significantly from the traditional Q-switched, solid state diode or lamp pumped, YAG technology.
Improvements over the prior art are desired with a method and system for generating pulses having a shape which is different than standard Q-switched pulses—pulses having faster rise time, relatively uniform and higher energy concentration in the central lobe, and fast fall time.